Some argue that the ROC plot adds no information beyond the AUC. Whether it adds, depends on how much you know about AUC/ROC and on what you want to know. I find them helpful. Here's why:
1. They hint if the sample size is sufficient
1. They hint if the sample size is sufficient
2. The differences between curves hint *how* adding predictors improves predictions. If the risk model will be used to identify high-risk indivs, the curves should differ in the lower left corner. Improvements only in the middle of the curve are unlikely relevant.
3. When the ROC curve follows the border of the plot, it means that sensitivity changes where specificity does not, or vice versa.
4. And sometimes it shows that the authors have no clue about the analyses they are doing ...
Plots and graphs are vizualizations of data, they illustrate what the numbers mean. The ROC plot is just a transformation of the risk distributions of people who did or did not develop the disease. Nothing more, nothing less.
/end
academic.oup.com/ije/advance-ar…
/end
academic.oup.com/ije/advance-ar…
